1. Technical Field
The present invention relates in general to process control systems. In particular, the present invention relates to process control optimization systems which utilize an adaptive optimization software system. In yet further particularity, the present invention relates to adaptive optimization software systems which comprise intelligent software objects (hereinafter sometimes "ISO" or "ISOs") arranged in a hierarchical relationship whereby the goal seeking behavior of each ISO can be modified by ISOs higher in the ISO's hierarchical structure. In yet further particularity, the present invention relates to ISOs comprising internal software objects including expert system objects, adaptive models objects, optimizer objects, predictor objects, sensor objects, and communication translation objects. In yet further and final point of particularity, the present invention also relates to a method of human interaction with said adaptive optimization software system.
2. Background Art
Process control systems are used in a variety of applications to sense process conditions and adjust process operating parameters in an attempt to optimize performance for given sets of goals. Many current conventional process control systems use static representations of the process to be controlled and do not provide for changes in the process control model being used in real time.
In conventional adaptive control theory, a suitable controller structure is chosen and the parameters of the controller are adjusted using static rules so that the output of the process follows the output of the reference of the model asymptotically. Static rules do not permit a process control system to automatically and optimally adapt to changing process conditions.
One significant deficiency of prior art process control systems, whether or not adaptive, is their lack of an intuitive user interface, either for initially configuring a system or for interacting with the system in real-time.
Another significant deficiency of prior art process control systems, whether or not adaptive, is the inability of the process control system to automatically perform control actions and, in so doing, provide a global goal-seeking mechanism that ties the process control system together into a powerful unified system to achieve the highest optimization congruent with management objectives and goals.
Further, many process control systems in the prior art provide for limited levels of control point, component, and/or system modeling or control hierarchies.
Accordingly, many prior art process control systems, whether or not adaptive, cannot provide concurrent multi-level optimization ranging from specific, component-oriented, narrowly focused levels to the broadest, global level.
Traditional process control systems are built up of discrete components (i.e., sensors, controllers) that work independently and lack low-level optimization. Some systems optimize on a global, system level without regard to optimization at each component level, while still other systems optimize only at the component level. As no global goal-seeking mechanism ties the parts together into a powerful unified system to achieve management objectives, the overall process fails to achieve its highest optimization and integration of low-level or component level optimization with the higher level or system level optimization.
Many systems that do provide some amount of concurrent multi-level optimization rely on just one or two methods of achieving the desired concurrent multi-level optimization, rather than on a multiplicity and variety of methods including expert systems, adaptive models which can use one or more modeling methodologies including neural networks, and other predictive modeling techniques.
Among the limited numbers of systems that use a variety of methods, no process control system uses interacting, differing adaptive methods to dynamically change its chosen predictive models in real-time without having to stop either the process being controlled or the process control system.
Moreover, many current conventional process control systems rely upon human operators to determine and implement optimum set points throughout the domain of the process control system in real-time. These process control systems require human intervention to optimize processes and systems, but because human operators vary greatly as to experience and the soundness of their control reasoning, this human factor introduces a wide-ranging variable in the overall effectiveness of the process control system.
Expert systems have provided a significant improvement over traditional process control systems that do not use expert systems. However, many current art process control systems do not use expert systems to assist in adaptation of process control algorithm operation, algorithm selection, or algorithm parameter estimation.
Further, once installed, current art process control systems that do use expert systems lack automatic, systematic approaches to adaptively optimizing its expert system and the expert system's algorithms.
Neural networks are a powerful modeling technique used to assure that the process model accurately predicts the performance of the modeled process over time. However, neural networks have a well known problem of "memorizing" and thereby becoming "static" and unable to find mutated, differing models to more accurately predict process performance over time.
Further, neural networks by definition depend on the user's omniscience to function correctly, and as user omniscience cannot be guaranteed, neural networks based systems suffer from reliance on user omniscience.
Moreover, some neural networks require weights used for the neural network's evaluation to be derived from both the constraints to be implemented and from any data functions necessary for solution; these may not be available as inputs to the network, thus limiting the neural networks' applicability to the process control system due to the inability to learn how to calculate these weights in real-time.
3. Disclosure of Invention
It is therefore an objective of the instant invention to provide a process control optimization system that uses dynamic representations of the process to be controlled, thus providing for changes in the process control model being used in real time.
It is a further objective of the instant invention to provide a process control optimization system that automatically and optimally adapts to changing process conditions.
It is a further objective of the instant invention to provide a process control optimization system having an intuitive user interface for both initially configuring a system and for interacting with the system in real-time.
It is a further objective of the instant invention to provide a process control optimization system having a global goal-seeking mechanism that ties a process control system together into a powerful unified system that achieves the highest optimization congruent with management objectives and goals.
It is a further objective of the instant invention to provide a process control optimization system having virtually unlimited levels of modeling and control hierarchy as well as virtually unlimited numbers of component-level process control points.
It is a further objective of the instant invention to provide a process control optimization system having concurrent multi-level optimization ranging from specific, component-oriented, narrowly focused levels to the broadest, global level.
It is a further objective of the instant invention to provide a process control optimization system that achieves the highest optimization and integration of low-level or component level optimization together with higher level or system level optimization.
It is a further objective of the instant invention to provide a process control optimization system having said concurrent multi-level optimization using a variety of methods including expert systems, adaptive models which can use one or more modeling methodologies including neural networks, and other predictive modeling techniques.
If is a further objective of the instant invention to allow a process control optimization system to use one or more interacting, differing adaptive methods to dynamically change its chosen predictive models in real-time without having to stop either the process being controlled or the process control system.
It is a further objective of the instant invention to provide a process control optimization system that optimizes processes and systems consistent with management objectives without the need for continuing human intervention.
It is a further objective of the instant invention to provide a process control optimization system that use expert systems to assist in adaptation of algorithm operation, algorithm selection, and algorithm parameter estimation
It is a further objective of the instant invention to provide a process control optimization system that automatically, systematically approaches adaptively optimizing its expert system and the expert system's algorithms.
It is a further objective of the instant invention to provide a process control optimization system that adaptively uses neural networks to prevent memorization by the neural networks.
It is a yet further and final objective of the instant invention to provide a process control optimization system that does not depend on a user's omniscience to function correctly.
As more fully described herein below, the present invention provides a process control optimization system that achieves, in substantial measure, these above stated objectives by including intelligent software objects (hereinafter referred to as an ISO or ISOs); an adaptive optimization software system comprising ISOs; a method of initializing said adaptive optimization software system; and a method of human interaction with said adaptive optimization software system.